Paste: z

Author: z
Mode: factor
Date: Thu, 20 Jul 2023 00:32:10
Plain Text | 
import tensorflow as tf

def smape_loss(y_true, y_pred):
    epsilon = 0.1  # 防止分母为0
    return 100 * tf.reduce_mean(2 * tf.abs(y_pred - y_true) / (tf.abs(y_pred) + tf.abs(y_true) + epsilon))

# 创建线性回归模型
model = LinearRegression()

# 编译模型，使用自定义的SMAPE损失函数
model.compile(optimizer='sgd', loss=smape_loss)

# 训练模型
model.fit(X_train_poly, y_train, epochs=100, batch_size=32, validation_split=0.1)

def smape_metric(y_true, y_pred):
    epsilon = 0.1  # 防止分母为0
    return 100 * tf.reduce_mean(2 * tf.abs(y_pred - y_true) / (tf.abs(y_pred) + tf.abs(y_true) + epsilon))

# 创建线性回归模型
model = LinearRegression()

# 编译模型，使用自定义的SMAPE损失函数和指标函数
model.compile(optimizer='sgd', loss=smape_loss, metrics=[smape_metric])

# 训练模型
model.fit(X_train_poly, y_train, epochs=100, batch_size=32, validation_split=0.1)

New Annotation

Summary:
Author:
Mode:
Body: